Control beliefs moderate emotion influences on complex ... and... · Control beliefs moderate...
Transcript of Control beliefs moderate emotion influences on complex ... and... · Control beliefs moderate...
COGNITION AND EMOTION, 2003 (under review)
Control beliefs moderate emotion influences
on complex problem solving
Miriam SperingRuprecht-Karls-Universität Heidelberg, Germany
Dietrich WagenerUniversität Mannheim, Germany
Joachim FunkeRuprecht-Karls-Universität Heidelberg, Germany
The assumption that positive affect leads to a better performance in various cognitive tasks hasbecome well established. We investigated whether positive and negative feedback-inducedemotions influence performance and strategies of 74 university students in a computer-simulated complex problem-solving scenario. We further analysed whether control beliefsmoderate the relation between emotions and complex problem solving. While overall scenarioperformance was not affected by emotions, strategy was: Participants with negative emotionswere more information-oriented in their problem-solving behaviour. Control beliefssignificantly interacted with induced emotions: Students with internal control beliefs benefitedmost from positive emotions, whereas students with external control beliefs performed bestwhen no emotions were induced. We suggest that the moderating effect may be due to amotivating side effect of the emotion elicitation and outline different approaches to test emotioninfluences in complex problem solving.
INTRODUCTION
While mathematician Andrew Wiles was brooding over Fermat’s last theorem – a
complex problem that had exasperated dozens of illustrious mathematicians before – he was in
a state of emotional agitation. In 1993, after living in seclusion for seven years, his first
thorough attempt of proof was heavily criticised. It was within this episode of distress and
depression that he brought about a cast-iron proof of Fermat’s last theorem (Singh, 1997). This
Emotions and complex problem solving 2
excerpt from the stirring history of Fermat’s last theorem shows that feelings can influence our
thinking in an either beneficial or harmful way. However, anecdotes of that kind confront us
with numerous questions: What emotions are beneficial to our problem-solving abilities and
what emotions impair our performance? In what tasks do emotions generally exert an influence
and why?
The present study addressed the effects of emotions on problem solving in complex and
dynamic situations. In recent years, there has been growing research interest in the role of affect
in cognitive processes (e.g., Fiedler, 2001; Schwarz, 2000). Yet, most empirical studies in this
field have focussed solely on unspecific positive or negative mood states and have employed
simple cognitive tasks rather than more complex problem-solving situations. With the current
investigation we pursued three goals. Firstly, we explored whether it is possible to induce
emotions in individuals who solve complex problems. Secondly, we tested whether the findings
from studies on the influence of affect on simple cognitive tasks apply to complex problems.
Since we expected that control beliefs play a distinct intervening role in emotion regulation, we
thirdly examined whether control beliefs moderate the influence of emotions on complex
problem solving. In accordance with these objectives, we clarify the three main issues of this
investigation in the following paragraphs and summarise our ideas in a schematic model.
THEORETICAL RATIONALE AND RESEARCH OVERVIEW
Complex problem solving: When you do not know what to do
As there is no unified definition of complex problem solving (CPS), the concept is often
circumscribed by contrasting it with simple cognitive tasks in terms of the following five
characteristic criteria: (a) complexity of the situation, (b) connectivity of variables, (c) dynamic
development, (d) intransparency or opaqueness, and (e) polytely (pursue of multiple goals; see
Frensch & Funke, 1995, for a detailed consideration).
Simple problem-solving tasks, e.g., Duncker's (1945) candle task, Maier's (1930) nine-
dot problem, or the famous Tower of Hanoi (e.g., Simon, 1975), require creative ingenuity and
restructuring of the given information. However, these tasks do not directly reflect the demands
of problems in complex and dynamic situations that people are confronted with in their daily
professional and private lives. In response to this criticism a new approach was initiated by
Broadbent (e.g., Broadbent, 1977) and Dörner (e.g., Dörner, 1980). Following this pioneering
Emotions and complex problem solving 3
work, research on CPS has made considerable headway in the last 30 years and has been
studied extensively by means of computer-simulated scenarios, like Lohhausen, Tailorshop, or
Biology Lab (see Funke, 2001, for an overview). These complex and semantically-rich
computerised tasks are constructed to mirror characteristics of real-life problems (e.g., Berry &
Broadbent, 1995; Brehmer, 1995; Dörner & Wearing, 1995). In the arguably best-known
scenario Lohhausen (Dörner, Kreuzig, Reither, & Stäudel, 1983), participants act as mayor of a
small town and have to direct the economic climate and people’s well-being to govern the town
successfully. These complex problem-solving tasks demand the gain and integration of
information, the elaboration and attainment of goals, the planning of action, decision-making,
and self-management (Dörner, 1986).
Up to now, most research on CPS centred around one of the three areas (a) the influence
of personal factors (e.g., intelligence, previous knowledge, or personality traits) on CPS, (b) the
influence of situational determinants (e.g., group problem solving, conceptual formulation, or
feedback) on CPS, and (c) system characteristics (e.g., formal aspects such as the semantic
embedding of the system). Despite considerable progress made in this research domain, the
question of whether affect (distinct emotions or moods) influences CPS has rarely been
addressed.
Emotions: When you know where your feelings come from
Looking at affective influences on cognition, it appears important to distinguish between
emotions and moods (e.g., Siemer, 2001). Emotions are commonly understood as short-lived,
intense phenomena that usually have a clear cognitive content and a salient cause that is highly
accessible to the person experiencing the emotion (e.g., Clore, Schwarz, & Conway, 1994).
Moods can be differentiated from emotions by their core properties (a) globality, (b)
diffuseness, (c) lack of intentionality, (d) longer duration, and (e) lower intensity (e.g., Clore et
al., 1994; Ekman, 1994; Frijda, 1994; Oatley & Johnson-Laird, 1987, 1996; Schwarz, 1990). As
described earlier, most studies to date have, by and large, focussed on the influence of moods
on cognitive processes or have only dealt with simple cognitive tasks. We summarise the results
of these studies and outline those investigations dealing with the influence of person variables
on complex problem-solving tasks. It should be noted that the terms affect, emotion, and mood
have been used interchangeably in most previous studies. Henceforth, we will employ these
terms in accord with the cited author.
Emotions and complex problem solving 4
Emotions and complex problem solving
A number of studies looked at the role of constructs such as perceived emotional
intelligence, emotional resilience, emotional reactivity, uncertainty, and anxiety as a trait in
complex problem-solving scenarios (Dörner, 1998; Stäudel, 1987). Likewise, other authors
have dealt with the influence of stress, coping abilities, and achievement motivation on
complex problem solving (Hesse, Spies, & Lüer, 1983). The results of these studies are
inconsistent: Whereas some authors have found emotional variables to impair complex problem
solving, most studies detected no effect at all. However, all studies have focussed almost solely
on the outcome of complex problem-solving tasks and disregarded the conceivable influence of
emotional variables on processing strategies. Moreover, emotional variables have not been
induced experimentally.
As previous research demonstrates, moods and emotions can profoundly influence both
strategic approaches and solution quality of cognitive tasks in general. Some of the uncertainty
surrounding the question of whether this influence is disruptive or facilitative seems now
resolved: The results of most studies, employing a wide range of induction methods and
cognitive tasks, indicate that positive affect enhances simple and creative problem solving (see
Isen, Daubman, & Nowicki, 1987). Positive affect also facilitates efficient decision making in
complex environments (e.g., in making a medical diagnosis; Isen, 2001). There is further
evidence for the contention that elated moods lead to flexible, creative, efficient, and thorough
thinking and that people prefer wider cognitive categories when induced with positive affect
(Bohner, Marz, Bless, Schwarz, & Strack, 1992; Fiedler, 2001; Forgas & Fiedler, 1996; Isen,
2001). In addition to that, affective states are assumed to influence the selection of strategies for
information processing in simple cognitive tasks. A large body of experimental research
documents that individuals in a positive affective state are more likely to use a heuristic
processing strategy (characterised by top-down processing and the productive use of generic
knowledge structures). Negative affect, on the other hand, facilitates careful bottom-up
processing and a more systematic gathering of information. Individuals in a bad mood pay more
attention to the details at hand (Fiedler, 2001; Hertel, Neuhof, Theuer, & Kerr, 2000; Schwarz,
2000).
Theoretical explanations. Different descriptive models have been offered for these
findings, as for instance the affect-infusion model (Forgas, 1994), the mood-and-general-
Emotions and complex problem solving 5
knowledge approach (Bless & Fiedler, 1995), or the affect-as-information hypothesis (Schwarz,
1990). Fiedler’s affect-cognition theory (2001) plausibly describes the underlying cognitive
processes and postulates that positive and negative mood states are affective cues to appetitive
and aversive settings: “While negative mood supports the conservative function of sticking to
the stimulus facts and avoiding mistakes, positive mood supports the creative function of active
generation, or enriching the stimulus input with inferences based on prior knowledge” (Fiedler,
2001, p. 3). Fiedler uses the Piagetian terms accommodation (tuning the cognitive system to fit
the stimulus environment) and assimilation (transforming external environment to fit internal
knowledge structures), to label these two processes, respectively. Negative mood thus
facilitates accommodation, and positive mood supports assimilative functions.
Control beliefs: When you either believe in yourself or trust in chance or others
CPS is not only thought to be influenced by emotional states but also by personality
factors, especially control beliefs. Generalised control beliefs constitute relatively stable
personality characteristics, reflecting an individual’s belief that he or she can cope with difficult
demands. This construct has been operationalised under a variety of different labels, namely
self-efficacy, optimistic self-belief, locus of control of reinforcement, and hope. According to
Krampen’s (1991) expectancy-value approach, control beliefs are theoretically derived from the
expectancy of contingency in action and represent the subjective expectancy that feasible
actions will be successful in any given situation. Similar to the notion of internal versus
external locus of control (LOC), Krampen conceives internal control beliefs as pointing to a
person’s opinion that he or she is in a position to influence events in any given situation by his
or her own action. External control beliefs indicate that an individual perceives upcoming
events as controlled by chance or other people’s influence. According to Krampen, people with
internal LOC are characterised by high activity, self-assuredness, inventiveness, quick decision-
making, and high responsibility. These individuals trust in their own abilities and feel relatively
independent of other persons or situations. People with external LOC, on the other hand, are
more insecure and passive when facing new situations. They can be distinguished by their low
self-assuredness and they feel dependent on other people or trust in chance and fate.
Control beliefs and complex problem solving
Previous studies on the relationship between personality factors and CPS have merely
concentrated on cognitive styles, e.g., impulsivity vs. reflexivity, strategic flexibility, or
Emotions and complex problem solving 6
heuristic competence, and on dispositional personality traits, such as extraversion, neuroticism,
and rigidity (Dörner et al., 1983; see Schaub, 2001, for an overview). The studies yielded a
variety of singular findings that cannot be fitted into a coherent pattern. Stäudel (1987) has
found positive correlations between problem-solving ability and self-assuredness. According to
Krampen (1991), self-assuredness is a component of internality in control beliefs and a relation
between control beliefs and CPS is therefore feasible. There are two ways in which control
beliefs may influence CPS: In addition to a direct effect, control beliefs could moderate the
relationship between emotions and CPS, as Dörner (1998) suggests that control beliefs affect
CPS by influencing the regulation of emotions (see also Bartl & Dörner, 1998). Individuals
with internality and externality in control beliefs should deal differently with emotions: We
expected that individuals with an internal LOC benefit more from positive emotions than do
individuals with an external LOC. Individuals with an external LOC should, however, be better
able to regulate negative emotions.
Figure 1. Schematic depiction of the interplay between emotions, control beliefs, complex problem-solving performance, and complex problem-solving behaviour. Arrows with labels H1 to H4 stand forthose hypotheses that were tested here.
The theoretical assumptions described so far are summarised in Figure 1. We provide
this as a heuristic framework to illustrate the theoretical context into which the current study is
incorporated. The framework depicts the origins and functions of emotions in a feedback loop.
Concerning the origin of emotions, a sensed value (evaluation) is compared to a reference value
Emotions and complex problem solving 7
(goal) that is set in accordance with control beliefs. If there is a positive discrepancy, positive
emotions arise; if the discrepancy is in fact negative, negative emotions are triggered (Carver &
Scheier, 1999). Those connections referring to the origin of emotions were not tested here. The
illustration shows the pertinent relations between control beliefs (as a trait), emotions, complex
problem-solving behaviour (action), and performance (outcome).
OVERVIEW OF THE PRESENT STUDY
We report and discuss the findings of an experimental study on the influence of
feedback-induced emotions and control beliefs as a trait on CPS. A computer-simulated
scenario was used to test complex problem-solving performance and problem-solving
behaviour. Four hypotheses were tested. Firstly, it is predicted that positive emotions, in
contrast to negative emotions, lead to a better performance in the complex problem-solving
scenario employed, as this task requires assimilative functions (Hypothesis 1). Secondly,
positive and negative emotions are predicted to lead to different problem-solving behaviours:
Positive emotions cause an intuitive, hypothesis-oriented approach, whereas negative emotions
are associated with more detail-oriented, information-based action (Hypothesis 2). Thirdly,
control beliefs are hypothesised to directly influence CPS. Individuals with internal LOC are
predicted to show a significantly higher problem-solving performance as compared to
participants with external LOC (Hypothesis 3). Finally, it is expected that control beliefs
moderate the influence of emotions on CPS. Individuals, who have an internal LOC, benefit
from positive emotions more than individuals with an external LOC. Likewise, individuals,
who are external in their control beliefs, perform better than individuals with internality in their
control beliefs when experiencing negative emotions (Hypothesis 4). Gender differences
(effects of biological sex) were also taken into account as an additional factor for each of the
hypotheses under study.
METHODS
Measures
Emotion induction. A wide variety of induction techniques have been employed in
emotion research amongst which music and films have been most successful (e.g., Niedenthal,
Emotions and complex problem solving 8
Halberstadt, & Setterlund, 1997; Westermann, Spies, Stahl, & Hesse, 1996). In selecting an
adequate treatment method for this study, two constraints had to be taken into account. Firstly,
the treatment had to be administered between the introductory period and the testing period of
the computer-simulated scenario and was hence restricted to a very short time frame. Secondly,
the emotion induction method had to be relevant to the problem-solving task to appear
authentic, which made music and films not useful as induction material. Positive and negative
emotions were instead elicited experimentally two times throughout the experiment by giving
false (either positive or negative) feedback on performance. The first feedback was
administered by telling students a false score after they had completed the spatial-reasoning
test, a subtest of the intelligence structure test 2000 (Intelligenz-Struktur-Test, I-S-T 2000;
Amthauer, Brocke, Liepmann, & Beauducel, 1999). The second feedback was provided after
half time of the computer-simulated scenario and was displayed automatically in a pop-up
window, which informed participants about their position in a fictitious ranking (positive: “You
are in position 12 out of 250 and are thus better than 95,2% of all participants”, negative: “You
are in position 208 out of 250. That means that 83,2% of all participants have performed better
than you”).
Emotion measures. Emotions were measured seven times using a 14-item questionnaire.
Participants were asked to mark a 10-cm line between the two poles null and all pervasive for
each of 14 emotion adjectives (content, sad, excited, tense, confused, angry, anxious, surprised,
enthusiastic, interested, calm, happy, ashamed, proud). Emotion labels were put in different
random order on each of the questionnaires to prevent sequence effects. This selection of
adjectives was based on a questionnaire by Schmidt-Atzert and Hüppe (1996) and on a German
version of Watson et al.’s PANAS scales (Krohne, Egloff, Kohlmann, & Tausch, 1996). The
questionnaire was a paper and pencil test and was announced by a pop-up window inside the
computer-simulated scenario.
Control beliefs. Krampen’s (1991) competence and control beliefs questionnaire
(Fragebogen zu Kompetenz- und Kontrollüberzeugungen, FKK) has been used frequently in a
variety of contexts since its first version was published in German (Krampen, 1981). The
questionnaire measures three aspects of LOC and one aspect of competence orientation, namely
internality, powerful others externality, chance control externality, and self-concept on four
distinguishable scales. These four scales can be aggregated to two secondary scales and a
Emotions and complex problem solving 9
tertiary scale (internality vs. externality in control beliefs) that will be used for purposes of data
analysis in this study. Krampen (1991) has argued that control beliefs can be domain-specific.
We therefore rephrased some of the items of the FKK, thus yielding a questionnaire measuring
control beliefs in problem solving, the FKK-PS. Table 1 shows four items and adjunctive scales
with indication of their reliability.
Table 1
Overview of FKK-PS scales and text examples with reliability indexes
Scale (with number of items) Item (with item-number) Cronbach’s alphaPrimary scales
„Whether I am able to solve a problemor not totally depends on my own effortand endurance.” (5)
.60
„Whether I am able to solve a problemor not depends on the support of otherpeople.” (26)
.74
„Whether I am able to solve a problemor not is entirely a matter of luck.” (15)
.72
Internality incontrol beliefs (8)
Social externality in controlbeliefs (8)
Chance control (8)
Self-concept of ownabilities (8)
„I sometimes do not have a clue what todo under the circumstances.“ (24)a
.78
Secondary scales.80Self-efficacy (16)
Externality in control beliefs (16) .78Tertiary scaleInternality versus externality incontrol beliefs (32)
.63
a Item is reverse coded.
Complex problem solving. The computer-simulated scenario FSYS 2.0 (Wagener, 2001)
was used in the current study1. In this dynamic and complex system, problem solvers have to
manage a forest enterprise over a period of 50 fictitious months (simulation cycles) at a profit.
In order to achieve that goal, they have to plant, raise, and fell trees in five structurally
equivalent partitions of the forest and have to maintain the wood’s quality (e.g., fertilise
grounds, do pest control). FSYS is based on Dörner’s (1986) model of demands on problem
solvers. The scenario is accordingly designed to indicate strategies in different aspects of
problem-solving behaviour additionally to the overall success in controlling the system. FSYS
delivers 14 scales that are depicted in Table 2. The scales are organised in four groups,
Emotions and complex problem solving 10
representing scenario performance (SP), quality of measures (QM), acquisition of information
(AI) and self-management (SM). For the 90-minute duration of the test session, the program
protocols and calculates the individual’s performance data with regard to the overall profit
yielded and behavioural components on the scales mentioned before.
Table 2
Overview of selected FSYS scales
Dimension and scale name Description of scaleScenario performance (SP)Earned capital (SKAPKOR) Account balance (income minus expenses for wood
maintenance) plus value of forest after 50 cyclesQuality of measures (QM)Prevention of errors (MNOFEHL)Assigning of priorities (MPRIORI)Efficiency of actions (MEFFIZI)Early comprehension (MVERSTA)
Correct dosage of e.g., fertiliser and pesticideAdequateness of decisions in goal conflictsTotal vs. partial achievement of subgoalsTime in scenario when all possible measures havebeen ordered once
Acquisition of information (AI)Early orientation (IORIENT)Information before acting(IVORHAND)Reading statistics (IFEEDS)
Surveying of forest areas (IKONTI)
Number of text elements accessed in first five cyclesKnowledge acquisition about effects of measuresbefore actingGaining overview of business process by readinggraphs and/or tablesContinuity of inspection of forest areas
Self-management (SM)Decidedness in action taking(ESICHER)
Number of decisions that have been cancelled withinthe same cycle
Additional measures. Personal data (e.g., age, gender, subject of studies) were collected
at the end of the experiment on a short standardised questionnaire that also asked for feedback
on the experiment and any inconveniences that might have disrupted a smooth procedure. No
such influences were reported.
Participants and procedure
Participants were 74 students and graduates of different fields of study; n = 32 were
female and n = 42 were male; the mean age was M = 24,6 (SD = 2,89). Students were contacted
in academic courses during March 2001 and took part on a voluntary basis. They were either
Emotions and complex problem solving 11
returned partial credit for a course requirement or got a book-gift. The experiment took place at
the Department of Psychology of the University of Heidelberg.
Participants were sent the modified version of the FKK-PS via email one week before
the experimental session and were asked to return the questionnaire before the session.
According to their result in the FKK-PS (either internality or externality in control beliefs)
respondents were assigned to three conditions (either positive or negative treatment or a control
group without any treatment) in a randomised manner thus resulting in a 2x3 design.
Individuals were distributed equally across the cells with n = 12 subjects in the four treatment
groups and n = 13 subjects in the two control groups. The students were then tested
individually. They first had to complete the spatial-reasoning test from the I-S-T 2000
(Amthauer et al., 1999) and were then given the written instructions for the computer-simulated
scenario. The space-test result was employed to induce emotions by giving false feedback on
the score. Emotions were measured directly before the space-test and after the feedback. People
then had 90 minutes to work on the scenario FSYS. Every tenth cycle, a pop-up window
instructed participants to complete another emotion questionnaire. After the 25th cycle, the
second feedback was given automatically in a pop-up window. Emotions were measured
immediately afterwards. Due to high standardisation of instructions and procedure, there was
no interaction between participants and experimenter except during the first feedback-cycle.
RESULTS
The significance level for all statistical hypotheses tests was set at p = .05; significance
levels for treatment-check purposes could be set at p = .01 since higher effect sizes were
expected here. All significance tests were one-tailed. In addition to levels of significance, effect
sizes d, f, or f2 are reported, following the conventions of Cohen (1987). An a priori power
analysis with G-Power (Erdfelder, Faul, & Buchner, 1996) revealed that, in consideration of a
sample size of N = 74, an expected large effect (f = 0.40) can be detected with a power of 0.85
for Hypotheses 1 and 2 and a medium effect (f2 = 0.15) can be verified with a power of 0.91 for
Hypothesis 3 and with a power of 0.84 for Hypothesis 4. When analyses of variance
(ANOVAs) are reported, different degrees of freedom within the same data analysis reflect the
fact that outliers were omitted.
Manipulation check: How good was the emotion elicitation?
Emotions and complex problem solving 12
To test whether the induction method was successful, emotion scores were compared
between the three groups for each of the two points of measurement before and after the
feedback was given. Emotion scores were derived from a differential R-technique factor
analysis (Barton, Cattell, & Curran, 1973) that yielded three factors. Equally weighted emotions
were aggregated to obtain three single emotion scores that represent the factors by summing
participants’ ratings of arousal, positive, and negative emotions. Subsequent one-way ANOVAs
were then carried out on these emotion scores with feedback as an independent variable. We
also conducted a multivariate analysis of variance (MANOVA) on emotion score, but because
the outcomes were comparable, we report the results of the ANOVA here. All treatment-group
means differed significantly from those of the control group in the expected direction.
Generally, positive feedback elicited substantially higher positive emotions whereas
negative feedback reduced positive emotions; F(2, 70) = 5.05, p < .01, f = 0.38 for the first
emotion induction, F(2, 71) = 10.03, p < .01, f = 0.53 for the second induction. Negative
emotions increased after negative feedback and decreased after the positive feedback,
respectively, but this effect was only significant for the second induction; F(2, 70) < 1, n.s., f =
0.12, F(2, 68) = 9.74, p < .01, f = 0.54. Treatment effects were not significant at any other point
of measurement after the first and second emotion induction. Concerning arousal, only the
second negative feedback as compared to the other conditions led to a stronger arousal, F(2, 71)
= 5.166, p < .01, f = 0.38.
To address the issue of which emotions were primarily elicited, effect sizes were
compared. The first and second positive induction chiefly elicited medium sized effects on
pride, while the two negative inductions increased anger and shame and led to reduced
contentment. Note that the effect sizes for shame (d = 0.57 and 0.80) and contentment (d = -
1.00 and -1.54) indicate medium to large effects for both, the first and the second induction,
respectively.
Finally, the manipulation check was conducted separately for women and men. Men
reacted significantly stronger to the second induction; F(2, 38) = 6.024, p < .01 for positive
emotions, F(2, 38) = 8.138, p < .01 for negative emotions. For women the effects of the second
induction were not significant; F(2, 27) = 2.510, n.s. for positive emotions, F(2, 27) = 2.552,
n.s. for negative emotions. As no gender differences were found for the first induction, this
result might be due to the fact that women were less involved in the scenario. A comparison of
Emotions and complex problem solving 13
means on the emotion scale interest yielded no significant differences, but there was a tendency
for reported means for women to gradually decrease whereas men’s interest remained stable.
To summarise, treatment effects were clearly demonstrated for the second emotion
induction. The effectiveness of the first induction with regard to negative emotions was less
clear than its effects on positive emotions. The effect sizes clearly show that there was a notable
effect on participants’ shame and contentment in both treatment groups.
Do emotions influence the outcome of complex problem solving?
To test the assumption that positive emotions facilitate and negative emotions impair
CPS (Hypothesis 1) under consideration of possible gender effects, a two-way ANOVA was
carried out on scenario performance with feedback and gender. The analysis showed that there
were no effects of emotions on the SP-scale (scenario performance), F(2, 68) < 1, n.s., f = 0.05.
However, a significant main effect of gender on scenario performance was detected, F(1, 68) =
4.763, p < .05. Male students performed better (M = 67.68, SD = 3.42) than did females (M =
56.23, SD = 3.98), d = 1.42.
Creative vs. careful processing: Do positive and negative emotions direct our strategies?
We tested whether positive and negative emotions lead to different information-
processing modes, as specified in Hypothesis 2. A MANOVA was computed on QM-scales
(quality of measures) and AI-scales (acquisition of information) with feedback as a factor.
Emotions indeed had an influence on the AI-scales that was significant for continuity in
information; F(2, 71) = 3.596, p < .05, f = 0.31. There was also a substantial effect for early
orientation in the scenario; F(2, 71) = 2.993, p = .05, f = 0.29. Participants with induced
negative emotions were more thorough in searching and using information as opposed to both
the control group and the positive treatment group. There were differences on the QM-scales
pointing to a better action handling in the group of positively induced participants but not
significantly so. The F-statistic for the multivariate test (Wilk’s lambda) was significant; F(14,
130) = 2.108, p < .05.
Control beliefs: Does trust in one’s own abilities lead to better performance?
In order to test the Hypothesis 3 that an internal LOC leads to a better performance in
CPS than an external LOC, a general linear model (GLM) on scenario performance with the
tertiary scale of the FKK-PS as an independent variable was carried out. Control beliefs had a
small effect (f2 = 0.03) on CPS. This effect was not significant, F(1, 72) = 2.024 – a finding
Emotions and complex problem solving 14
which might, in fact, be due to sample size and power (a post-hoc analysis revealed a power of
1 - b = 0.31).
Do control beliefs moderate the influence of emotions on complex problem solving?
Next, the moderator hypothesis (Hypothesis 4) was addressed. A GLM with treatment
(positive, negative, no treatment), control beliefs (as measured by the tertiary scale of the FKK-
PS), and the interaction between feedback and control beliefs revealed a significant interaction
between emotion and control beliefs; F(2, 68) = 3.390, p < .05. The effect can be considered as
medium sized with f2 = 0.15.
Figure 2. Interaction between treatment (positive, negative, or control group) and control beliefs(internal or external), showing that control beliefs moderate the influence of emotions on scenarioperformance.
A median split was done on the tertiary scale for illustration purposes to demonstrate the
interaction between feedback and control beliefs. The results of this analysis are depicted in
Figure 2. It can be seen that participants with an internal LOC performed best after receiving an
achievement feedback, especially if this was a positive feedback. They were also better able to
regulate negative emotions. However, participants with an external LOC performed best when
no emotion was induced. These results suggest that the feedback might have had a motivating
side effect.
Emotions and complex problem solving 15
To take possible gender differences into consideration, separate GLMs by gender were
computed. Surprisingly, the interaction between feedback and control beliefs existed for men,
F(2, 36) = 5.750, p < .01, but not for women, F(2, 26) < 1, n.s.. To summarise, control beliefs
can indeed be seen as a moderator variable that contributes to the relation between emotions
and CPS. The role of gender and the feedback’s motivating effect, which might, in fact, be
responsible for the significance of the interaction, will be discussed below.
DISCUSSION
How did Andrew Wiles so successfully prove Fermat’s last theorem apart from being a
mathematical genius? The answer to that question is not clear based upon the results presented
here, since his task was different from the scenario undertaken by our subjects. However, our
results clearly demonstrate that emotions influence complex problem-solving behaviour,
although they do not exert an influence on overall scenario success unless control beliefs are
considered as a moderator variable.
Emotions do not influence complex problem solving
Concerning the influence of positive and negative emotions on scenario performance
(Hypothesis 1), there were no significant differences between participants that were induced
with positive emotions and those students that received a negative treatment. This result is
inconsistent with most previous studies on simple problem solving (e.g., Isen, 2001). To
account for the finding, we will concentrate on two possible causes: (a) the conceivable
inefficiency of the treatment, and (b) the nature and demands of the complex problem-solving
situation.
Firstly, the result could originate from an inefficient treatment, but the manipulation
check demonstrated that both treatments elicited emotions, and that there were medium to large
effects on the emotion dimensions. Three other issues are related to the question of treatment
efficiency: (a) the intensity and duration of evoked emotions, (b) the influence of labelling state
emotions, and (c) the attention that individuals pay to affective cues and the perceived
relevance of emotions. Considering the first point, differences in emotions were only significant
directly after the treatment was given. Apparently, induced emotions have been intense but
short and did not last long enough to cause substantial differences in scenario performance. As
to the second point, there is some evidence that labelling state emotions can reduce their impact
Emotions and complex problem solving 16
on cognitive processes (Keltner, Locke, & Audrain, 1993). Emotions have been measured
seven times throughout the experiment in order to gain information about the duration of the
treatment effects and the stability of emotions. Based on the findings of Keltner et al. (1993),
we do assume that the number of emotion questionnaires applied in this study could have
caused an unwanted reduction of emotion influence on measurable cognitive processes.
Concerning attention to emotions, Gasper and Clore (2000) found that the perception of the
relevance of affective cues influences the use of emotions as a basis for judgement. According
to the affect-as-information hypothesis (Schwarz, 1990), affect influences cognitive processes
only when it is experienced as a source of relevant information. We presume here that
participants valued emotional states as relevant, because emotions were induced by
performance feedback directly relevant to the task at hand. Yet, we did not control whether and
how often individuals actually attended to their emotions and can thus only suspect that some
students might have rarely noticed their feelings, let alone considered their feelings as a useful
source of information.
Three conclusions can be drawn from these considerations. First, other emotion
elicitation methods (i.e., inducement by music, pictures, or odours) might well prove to be more
efficient. The music method, however, is only feasible when it does not interfere with the
requirements of the task, as would have been the case here. Alternatively, performance
feedback could have been applied after each cycle. Problematic as this might be, a sustained
treatment could also prohibit effects of scenario-inherent feedback (such as graphical process
information) on students’ performance. Second, the efficiency and duration of the treatment
should be tested beforehand and questionnaires should be omitted in the main experiment to
prevent a possible reduction of effects. Third, a variety of measures have been developed to
assess individual differences in emotional attention and monitoring (for an overview see Gasper
& Clore, 2000; Otto, Döring-Seipel, Grebe & Lantermann, 2001). We think it advisable to
measure emotional attention and clarity as well as emotional stability as traits, especially in a
demanding complex problem-solving situation.
This leads us to a second possible explanation for the findings to Hypothesis 1 that
relates to the nature of the task employed here. The computer-simulated scenario FSYS
constitutes a highly complex, challenging and intriguing situation for the problem solver.
Emotions and complex problem solving 17
Students might have felt required to regulate their emotions or have been focussing more on the
cognitive task than on their feelings.
Individuals with negative emotions are well-informed problem solvers
Concerning Hypothesis 2, it was shown that positive and negative emotions elicited
distinguishable problem-solving strategies: As predicted by Fiedler’s (2001) affect-cognition
model, negative emotions led to a more detailed information search and to a more systematic
approach to the scenario. The differences on the scale for early orientation in the first five
months can be assumed to be a direct cause of the first treatment. As far as the effect of
treatment on complex problem-solving behaviour is concerned, it can be concluded that
Fiedler’s model applies well to emotions in CPS.
The results of Hypothesis 2 also show that the treatment has been successful. It can be
inferred that emotions do, in fact, influence CPS on the levels of action handling and
information acquisition. One possible conclusion is that different emotion-induced strategies for
CPS lead to the same result in the scenario FSYS. On the one hand, the problem solver has to
be well informed and can thus benefit from negative emotions. On the other hand, quick and
flexible decision-making is required – an approach that is basically elicited by positive emotion.
It is, in fact, probably the switching between different strategies (elicited by both, positive and
negative emotions) that accounts for success in complex problem solving. From this assumption
we draw the conclusion that the dichotomy between positive and negative emotions does not
necessarily lead to a classification of successful and unsuccessful problem solvers in a complex,
dynamic scenario.
Control beliefs do only play a marginal role in complex problem solving
Concerning the role of control beliefs (Hypothesis 3), our aim was twofold: It was
hypothesised that individuals with an internal LOC should be better problem solvers than those
participants with an external LOC. Astonishingly, there were no such differences between
problem solvers with internality versus externality in control beliefs, at least not significantly
so. However, a small effect was found. This finding is consistent with the results yielded from
previous work (e.g., Stäudel, 1987), stating that control beliefs play a marginal role in CPS.
Yet, control beliefs appear to moderate the effect of emotions on CPS, as the results to
Hypothesis 4 indicate: Participants with internal control beliefs benefited most from positive
emotions. Individuals with external control beliefs, however, did not perform best when
Emotions and complex problem solving 18
negative emotions were elicited, but when no emotions were induced. In contrast to our
expectations, we did not find significant differences between the two treatment groups and the
control group, but between those students who received a feedback and the control group.
Considering the theoretical background of the construct of control beliefs (e.g., Krampen,
1991), we claim that the feedback had a motivating effect on those students with an internal
LOC and that the interaction between treatment and control beliefs is thus due to the
motivational effects of the feedback. Being more active and achievement oriented, individuals
with internal control beliefs can be assumed to seek other people’s feedback on certain tasks
that are relevant to them. In contrast what is suggested by previous work, we conclude that
individuals with an internal orientation in control beliefs depend on external acknowledgement
of their performance to be successful. Participants with an external LOC, on the other hand,
might have mistrusted the positive performance feedback since they hold luck or other people
responsible for their own success or failure.
The role of gender: Men and women perform significantly differently
Although there is evidence in the literature for gender differences in cognitive tasks, we
did not expect to find any differences here since FSYS was formerly believed to be a gender-
neutral scenario (Wagener, 2001). In addition to the main effect of gender on scenario
performance, the moderator effect of control beliefs only holds for men: For women, there was
no significant interaction between control beliefs and emotions. The wide difference between
both groups with regard to self-reported positive and negative emotions was also astonishing,
especially since the literature does, succinctly, not indicate any consistent gender differences in
emotionality (Brody & Hall, 2000). The results of the manipulation check show men to be more
reactive to the emotion induction, especially to the second treatment. Furthermore, men seemed
to be more involved in the scenario, as the comparison of means on the subscale interest
indicates. Women, on the other hand, were less reactive to the feedback. How can we account
for these gender differences? The differences in reactivity to the feedback might simply be due
to the fact that women were more emotionally stable or less in need for a feedback. However, it
is more likely that women were less involved in the scenario and consequently less interested in
their personal success and therefore less exerted.
OUTLOOK
Emotions and complex problem solving 19
The findings of the present study have interesting implications on both empirical and
theoretical levels. Many aspects of the results demonstrate that studying emotions within the
research domain of CPS requires a different approach from analysing the effects of induced
emotions on simple cognitive processes. Future research should therefore concentrate on (a) the
formation and effect of different problem-solving strategies on behaviour in complex tasks
rather than on emotion-influences on overall success, (b) the moderating influence of traits,
e.g., control beliefs or emotional intelligence, on complex problem solving. Considering the
attention to, clarity, and repair of emotions as moderating trait aspects (Otto et al., 2001) might
enhance the understanding of emotion influences on cognitive processes in complex and
dynamic situations. Moreover, as the affect-cognition link is not unidirectional, we can assume
that perceived success or failure in FSYS might have triggered emotions at any time throughout
the scenario. Either, emotions have to be induced each cycle or post-cognitive emotions have to
be controlled by means of process-tracing methods (thus testing the assumed relations in Figure
1). Several attempts have already been made to compare precise emotions of the same valence
(e.g., Lerner & Keltner, 2000) and further endeavour should be made to induce emotions
purposefully (such as shame and pride by using performance feedback).
It also seems worth to consider extending other theoretical approaches from cognitive
and social psychology to CPS. Mellers has offered a decision-affect theory of anticipated
emotions that takes the emotional responses to the outcome of choice into account (e.g.,
Mellers, Schwartz, & Ritov, 1999). Decision-affect theory provides a descriptive account of
actual emotions and would thus allow us to examine connections between anticipated emotions,
strategy selection in CPS, and actual emotional responses. It might also be possible to
differentiate between distinct emotions in order to answer the normative question whether
people should actually employ emotions as a relevant parameter to guide their behaviour in
complex situations.
Emotions and complex problem solving 20
REFERENCES
Amthauer, R., Brocke, B., Liepman, D., & Beauducel, A. (1999). Intelligenz-Struktur-Test2000 [Intelligence Structure Test 2000]. Göttingen, Germany: Hogrefe.
Bartl, C. & Dörner, D. (1998). PSI: A theory of the integration of cognition, emotion, andmotivation. In F.E. Ritter & R.M. Young (Eds.), Proceedings of the Second EuropeanConference on Cognitive Modeling (pp. 66-73). Nottingham, England: NottinghamUniversity Press.
Barton, K., Cattell, R.B., & Curran, J.P. (1973). Psychological states: Their definition throughP-technique and Differential R (dR) technique factor analysis. Journal of BehavioralScience, 1, 273-278.
Berry, D.C. & Broadbent, D.E. (1995). Implicit learning in the control of complex systems. InP.A. Frensch & J. Funke (Eds.), Complex problem solving: The European perspective(pp. 131-150). Hillsdale, NJ: Lawrence Erlbaum Associates.
Bless, H. & Fiedler, K. (1995). Affective states and the influence of activated generalknowledge. Personality and Social Psychology Bulletin, 21, 766-778.
Bohner, G., Marz, P., Bless, H., Schwarz, N., & Strack, F. (1992). Zum Einfluß vonStimmungen auf Attributionsprozesse [On the influence of mood on attributionprocesses]. Zeitschrift für Sozialpsychologie, 23, 194-205.
Brehmer, B. (1995). Feedback delays in complex dynamic decision tasks. In P.A. Frensch & J.Funke (Eds.), Complex problem solving: The European perspective (pp. 103-130).Hillsdale, NJ: Lawrence Erlbaum Associates.
Broadbent, D.E. (1977). Levels, hierarchies, and the locus of control. Quarterly Journal ofExperimental Psychology, 29, 181-201.
Brody, L.R. & Hall, J.A. (2000). Gender, emotion, and expression. In M. Lewis & J. Haviland-Jones (Eds.), Handbook of emotions (pp. 338-349). New York: Guilford.
Carver, C.S., & Scheier, M.F. (1999). Origins and functions of positive and negative affect: Acontrol-process view. Psychological Review, 97, 19-35.
Clore, G.L., Schwarz, N., & Conway, M. (1994). Affective causes and consequences of socialinformation processing. In R.S. Wyer & T.K. Srull (Eds.), Handbook of social cognition(pp. 323-417). Hillsdale, NJ: Lawrence Erlbaum Associates.
Cohen, J. (1987). Statistical power analysis for the behavioral sciences (Rev. ed.). Hillsdale,NJ: Lawrence Erlbaum Associates.
Dörner, D. (1980). On the difficulties people have in dealing with complexity. Simulation &Games, 11, 87-106.
Dörner, D. (1986). Diagnostik der operativen Intelligenz [Diagnosis of operative intelligence].Diagnostica, 32, 290-308.
Dörner, D. (1998). Emotionen, kognitive Prozesse und der Gebrauch von Wissen [Emotions,cognitive processes, and use of knowledge]. In F. Klix & H. Spada (Eds.), Enzyklopädieder Psychologie, Kognition, Bd. 6: Wissen (pp. 301-335). Göttingen, Germany: Hogrefe.
Dörner, D., Kreuzig, H.W., Reither, F., & Stäudel, T. (1983). Lohhausen: Vom Umgang mitUnbestimmheit und Komplexität [Lohhausen: On handling indefiniteness andcomplexity]. Bern, Switzerland: Huber.
Dörner, D. & Wearing, A.J. (1995). Complex problem solving: Toward a (computersimulated)theory. In P.A. Frensch & J. Funke (Eds.), Complex problem solving: The Europeanperspective (pp. 65-99). Hillsdale, NJ: Lawrence Erlbaum Associates.
Emotions and complex problem solving 21
Duncker, K. (1945). On problem solving. Psychological Monographs, 58 (270).Ekman, P. (1994). Moods, emotions, and traits. In P. Ekman & R.J. Davidson (Eds.), The
nature of emotion. Fundamental questions (pp. 411-430). New York: Oxford UniversityPress.
Erdfelder, E., Faul, F., & Buchner, A. (1996). GPOWER: A general power analysis program.Behavior Research Methods, Instruments, & Computers, 28, 1-11.
Fiedler, K. (2001). Affective states trigger processes of assimilation and accomodation. In L.L.Martin & G.L. Clore (Eds.), Theories of mood and cognition: A user's guidebook (pp. 85-98). Mahwah, NJ: Lawrence Erlbaum Associates.
Forgas, J.P. (1994). The role of emotion in social judgements: An introductory review and anaffect infusion model (AIM). European Journal of Social Psychology, 24, 1-14.
Forgas, J.P. & Fiedler, K. (1996). Mood effects on intergroup discrimination: The role of affectin reward allocation decisions. Journal of Personality and Social Psychology, 70, 28-40.
Frensch, P.A. & Funke, J. (1995). Definitions, traditions, and a general framework forunderstanding complex problem solving. In P.A. Frensch & J. Funke (Eds.), Complexproblem solving: The European perspective (pp. 3-25). Hillsdale, NJ: Lawrence ErlbaumAssociates.
Frijda, N.H. (1994). Emotions require cognitions, even if simple ones. In P. Ekman & J.Davidson (Eds.), The nature of emotion: Fundamental questions (pp. 197-212). NewYork: Oxford University Press.
Funke, J. (2001). Dynamic systems as tools for analysing human judgment. Thinking andReasoning, 7, 69-89.
Gasper, K. & Clore, G.L. (2000). Do you have to pay attention to your feelings to be influencedby them? Personality and Social Psychology Bulletin, 26, 698-711.
Hertel, G., Neuhof, J., Theuer, T., & Kerr, N.L. (2000). Mood effects on cooperation in smallgroups: Does positive mood simply lead to more cooperation? Cognition and Emotion,14, 441-472.
Hesse, F.W., Spies, K., Lüer, G. (1983). Einfluß motivationaler Faktoren auf dasProblemlöseverhalten im Umgang mit komplexen Problemen [Influence of motivationalfactors on performance in complex problem solving]. Zeitschrift für Experimentelle undAngewandte Psychologie, 30, 400-424.
Isen, A.M. (2001). An influence of positive affect on decision making in complex situations:Theoretical issues with practical implications. Journal of Consumer Psychology, 11, 75-85.
Isen, A.M., Daubman, K.A., & Nowicki, G.P. (1987). Positive affect facilitates creativeproblem solving. Journal of Personality and Social Psychology, 52, 1122-1131.
Keltner, D., Locke, K.D., & Audrain, P.C. (1993). The influence of attributions on therelevance of negative feelings to personal satisfaction. Personality & Social PsychologyBulletin, 19, 21-29.
Krampen, G. (1981). IPC-Fragebogen zu Kontrollüberzeugungen [IPC questionnaire on controlbeliefs]. Göttingen, Germany: Hogrefe.
Krampen, G. (1991). Fragebogen zu Kompetenz- und Kontrollüberzeugungen (FKK)[Questionnaire on competence beliefs and control beliefs (FKK)]. Göttingen, Germany:Hogrefe.
Krohne, H.W., Egloff, B., Kohlmann, C.-W., & Tausch, A. (1996). Untersuchungen mit einerdeutschen Version der "Positive and Negative Affect Schedule" (PANAS) [Investigations
Emotions and complex problem solving 22
with a German version of the Positive and Negative Affect Schedule (PANAS)].Diagnostica, 42, 139-156.
Lerner, J.S. & Keltner, D. (2000). Beyond valence. Toward a model of emotion-specificinfluences on judgement and choice. Cognition and Emotion, 14, 473-493.
Maier, N.R.F. (1930). Reasoning in humans: I. On direction. Journal of ComparativePsychology, 10, 115-143.
Mellers, B., Schwartz, A., & Ritov, I. (1999). Emotion-based choice. Journal of ExperimentalPsychology: General, 128, 332-345.
Niedenthal, P.M., Halberstadt, J.B., & Setterlund, M.B. (1997). Being happy and seeing"happy": Emotional state mediates visual word recognition. Cognition and Emotion, 11,403-432.
Oatley, K. & Johnson-Laird, P.L. (1987). Towards a cognitive theory of emotion. Cognitionand Emotion, 1, 29-50.
Oatley, K. & Johnson-Laird, P.L. (1996). The communicative theory of emotions: Empiricaltests, mental models, and implications for social interaction. In L.L. Martin & A. Tesser(Eds.), Striving and feeling: Interactions among goals, affect, and self-regulation (pp.363-380). Mahwah, NJ: Lawrence Erlbaum Associates.
Otto, J.H., Döring-Seipel, E., Grebe, M., & Lantermann, E.-D. (2001). Entwicklung einesFragebogens zur Erfassung der wahrgenommenen emotionalen Intelligenz:Aufmerksamkeit auf, Klarheit und Beeinflussbarkeit von Emotionen [Development of aquestionnaire for measuring perceived emotional intelligence: Attention to, clarity, andrepair of emotions]. Diagnostica, 47, 178-187.
Schaub, H. (2001). Persönlichkeit und Problemlösen: Persönlichkeitsfaktoren als Parametereines informationsverarbeitenden Systems [Personality and problem solving: Personalityfactors as parameters of an information processing system]. Weinheim, Germany:Psychologie Verlags Union.
Schmidt-Atzert, L. & Hüppe, M. (1996). Emotionsskalen EMO 16. Ein Fragebogen zurSelbstbeschreibung des aktuellen emotionalen Gefühlszustandes [Emotion scales EMO16. An inventory for the self-rating of current emotional states]. Diagnostica, 42, 242-267.
Schwarz, N. (1990). Feelings as information: Informational and motivational functions ofaffective states. In E.T. Higgins & R.M. Sorrentino (Eds.), Handbook of motivation andcognition (pp. 527-561). New York: Guilford Press.
Schwarz, N. (2000). Emotion, cognition, and decision making. Cognition and Emotion, 14,433-440.
Siemer, M. (2001). Mood-specific effects on appraisal and emotion judgements. Cognition andEmotion, 15, 453-485.
Simon, H.A. (1975). The functional equivalence of problem solving skills. CognitivePsychology, 7, 268-288.
Singh, S. (1997). Fermat's last theorem. The story of a riddle that confounded the world'sgreatest minds for 358 years. London: Fourth Estate.
Stäudel, T. (1987). Problemlösen, Emotionen und Kompetenz: Die Überprüfung einesintegrativen Konstrukts [Problem solving, emotions, and competence: Examination of anintegrative construct]. Regensburg, Germany: Roderer.
Wagener, D. (2001). Psychologische Diagnostik mit komplexen Szenarios [Psychologicaldiagnosis with complex scenarios]. Lengerich, Germany: Pabst Science Publishers.
Emotions and complex problem solving 23
Westermann, R., Spies, K., Stahl, G., Hesse, F.W. (1996). Relative effectiveness and validity ofmood induction procedures: A meta-analysis. European Journal of Social Psychology, 26,557-580.
Author Note
Miriam Spering, Institute of Psychology, Ruprecht-Karls-Universität Heidelberg; Dietrich
Wagener, Institute of Psychology, Universität Mannheim; Joachim Funke, Institute of
Psychology, Ruprecht-Karls-Universität Heidelberg.
Correspondence concerning this article should be addressed to Miriam Spering, Institute
of Psychology, Hauptstr. 47-51, 69117 Heidelberg. E-mail: [email protected]
heidelberg.de